What is Cowork? Use Cases Blog Services AI Readiness Quiz FAQ Book a Call

Blog · Compliance · 2026

Can RIAs and broker-dealers use AI tools without triggering SEC issues?

A 2026 field guide to Form ADV disclosures, the Marketing Rule, Reg BI, Books and Records, and Reg S-P for US financial advisors deploying generative AI in their practice.

The short answer

Yes — RIAs and broker-dealers can use generative AI tools responsibly today, provided the firm treats AI output as draft work product reviewed by a qualified supervised person, retains AI-assisted client communications under Advisers Act Rule 204-2 (or Exchange Act Rule 17a-4 for broker-dealers), updates the WISP and Reg S-P safeguards to cover any AI vendor handling customer records, and avoids delegating the fiduciary or Reg BI duty of care to a model. The SEC has not banned AI. What it expects is that existing rules apply.

The regulatory backdrop

SEC AI-related rulemaking, 2023 to 2025

Three strands of SEC activity shape the 2026 picture:

  • Proposed predictive data analytics rule (July 2023). The Commission proposed Rule 211(h)(2)-4 under the Advisers Act and parallel broker-dealer rules, targeting conflicts of interest where covered technology is used in investor interactions. The proposal was criticized for scope and remains unfinalized in its original form as of 2026, but the staff position is clear: conflicts arising from AI use must be identified, eliminated or neutralized under existing fiduciary principles.
  • Marketing Rule interpretations and Risk Alerts. The Division of Examinations has published multiple Risk Alerts clarifying how Rule 206(4)-1 applies to AI-generated marketing, performance presentations, hypothetical performance, and testimonials. The 2024 and 2025 exam priorities both flagged AI washing — overstated claims about AI capabilities — as an enforcement focus.
  • Reg S-P amendments (May 2024). Final amendments imposed incident-response and client notification requirements for breaches of sensitive customer information, including breaches at service providers. AI vendors are service providers.

Treat these as the regulatory floor. Your state securities administrator may expect more.

Form ADV Part 2 disclosures: do you need to tell clients?

Form ADV Part 2A is where material facts about advisory business and methods live. The disclosure question for AI turns on materiality:

  • Item 4 (advisory business). If AI is central to how you deliver services — for example, you market algorithmic model portfolios or an AI-driven planning engine — describe it here.
  • Item 8 (methods of analysis, investment strategies, risk of loss). If AI meaningfully informs recommendations or portfolio decisions, disclose the methodology and its limitations, including model risk and the possibility of inaccurate outputs.
  • Item 10 and Item 14. If AI vendors create conflicts of interest or affiliate relationships, address them.

Internal productivity use — drafting a client letter that a licensed advisor reviews before sending, summarizing a meeting into CRM notes — is generally not a material business practice. The safer posture is to document the firm's position in writing, update the brochure at the next annual amendment, and have counsel sign off.

The Marketing Rule: AI, performance claims, and testimonials

Rule 206(4)-1 applies to any communication to more than one person that offers advisory services. Four pressure points for AI:

  • Content review. Every AI-drafted marketing piece requires pre-use review by a qualified person under the firm's written policies. The Marketing Rule is a strict-liability regime for misleading content.
  • Performance presentations. If AI generates or formats performance figures, the firm is on the hook for compliance with net-of-fees, time-period, and prescribed-metric requirements. Hypothetical performance obligations apply if AI projects returns.
  • Testimonials and endorsements. AI-generated reviews passed off as organic client testimonials are a clear violation. Disclosures about compensation and material conflicts apply.
  • AI washing. Marketing that overstates how AI is actually used — claiming proprietary AI when you use a general tool — is actionable under the general prohibitions.

Reg BI: AI-assisted suitability for broker-dealers

Regulation Best Interest's Care Obligation requires the registered representative to exercise reasonable diligence, care, and skill. SEC staff bulletins on standards of conduct repeatedly emphasize that technology cannot transfer the duty of care from the licensed rep to a machine.

In practice: AI can summarize a client's investment profile, structure a rationale memo, compare product features, and draft the explanation of the recommendation. The rep must independently evaluate suitability, document the reasoning, and sign off. Firms should update WSPs to describe the review step, log AI assistance in the recommendation file, and train supervisors on reviewing AI-informed recommendations.

Books and Records: retaining AI-assisted communications

Advisers Act Rule 204-2 generally requires a five-year retention of records relating to recommendations, advice, and client communications, with the first two years readily accessible. Exchange Act Rule 17a-4 imposes a six-year framework for broker-dealers.

  • Retain the final client-facing output — letter, email, memo, meeting summary — as you would any client communication.
  • Where AI materially shaped a recommendation, retain enough prompt context to reconstruct the basis of advice.
  • Store in a WORM-compliant or equivalent system where 17a-4 applies.
  • Capture AI chat logs from firm-approved tools through the firm's archiving solution. Shadow IT — personal accounts on consumer AI products — is a records violation waiting to happen.

Reg S-P: client data in AI tools and the WISP

Regulation S-P requires written policies and procedures to safeguard customer records and information. The 2024 amendments added incident-response, oversight of service providers, and client notification obligations. For AI:

  • Any AI tool handling customer PII must be inventoried in the WISP and covered by vendor due diligence.
  • A written data processing addendum should address training on inputs (prefer tools where the vendor does not train on your prompts), subprocessors, data residency, retention, and deletion.
  • Incident-response plans must contemplate AI vendor breaches, with the 30-day client notification clock starting when the firm becomes aware a breach likely occurred.
  • Many firms anonymize client identifiers before inputting to AI as an additional control, even when the DPA is solid.

A practical five-step compliance framework

  1. Inventory. List every AI tool in use at the firm, including consumer tools employees may be using informally. You cannot govern what you do not know about.
  2. Approve a stack. Select firm-approved tools with enterprise terms, a signed DPA, and no training on your inputs. Prohibit everything else in writing.
  3. Update policies. Amend the WISP, Code of Ethics, advertising and marketing procedures, and books-and-records policy to address AI. Address retention, review, prohibited uses (unsupervised client-facing output, performance projections without compliance review), and incident response.
  4. Train and attest. Train supervised persons on the policy, document attendance, and collect annual attestations. Record training materials for the file.
  5. Supervise and test. Add AI-use testing to the annual Rule 206(4)-7 compliance review. Sample AI-assisted client communications, confirm reviewer sign-off, verify records are captured, and update the risk assessment.

Firms that complete these five steps address the overwhelming majority of SEC examiner expectations for 2026.

FAQ

Common questions from US advisory firms

Do RIAs need to disclose AI use in Form ADV Part 2?

If AI is material to how you deliver advice, how you analyze portfolios, or how you communicate with clients, disclosure in Item 8 (methods of analysis) and potentially Item 4 (business) is prudent. Internal productivity use that a supervised person reviews before anything reaches a client is generally not a material business practice requiring disclosure, but firm counsel should make the call.

Can we use AI to draft marketing content under the Marketing Rule?

Yes, provided every AI-assisted piece is reviewed and approved by a qualified person before use, performance claims comply with Rule 206(4)-1 presentation requirements, no AI-generated testimonial or endorsement is passed off as organic, and records are retained under Rule 204-2. AI is a drafting tool; the Marketing Rule obligations sit with the firm.

What about Reg BI and AI-assisted suitability?

For broker-dealers, Reg BI's Care Obligation requires reasonable diligence by the registered representative. AI can summarize client profiles and draft rationale memos, but the rep must independently assess suitability. SEC staff bulletins emphasize that technology does not transfer the duty of care.

How long do we retain AI chat logs?

Advisers Act Rule 204-2 generally requires five-year retention of communications relating to recommendations, advice, or receipt of funds. If an AI chat generated content used with a client, the output and relevant prompt context should be retained in a readily accessible format for the same period. Broker-dealers follow Exchange Act Rule 17a-4's six-year framework.

Can client PII go into an AI tool under Reg S-P?

Reg S-P requires written policies to safeguard customer records. If you input client PII into an AI tool, the vendor must be covered by your WISP, vendor due diligence, and a written data processing addendum. Many firms anonymize client identifiers as an added control. The amended Reg S-P incident-response and notification requirements also apply to AI vendors.

Does the SEC's proposed predictive data analytics rule apply to generative AI?

The 2023 proposal targeted conflicts of interest arising when advisers use covered technology in investor interactions. The scope would reach some generative AI uses, particularly where outputs influence investment recommendations. As of 2026 the rule is not final in its original form, but firms are expected to identify and address conflicts proactively under existing fiduciary principles.

More questions? Read the full FAQ or book a call.

Build an AI-ready compliance program

Claudable trains US RIAs and broker-dealer practices to use Claude Cowork under a documented, examiner-ready policy — so the hours saved on drafting do not become hours spent in the next exam.

See Training Options Book a Discovery Call

From $199 per person · Team (up to 5) $449 · Implementation from $2,500